Re: [PATCH v2 3/4] mm/mmap: Change detached vma locking scheme

[Liam R. Howlett]

 Hello Andrew,

Please replace v2 with the attached v3 of this patch to address the
issue with building ARCH=um [1].

[1] https://lore.kernel.org/linux-mm/20230718172105.GA1714004@dev-arch.thelio-3990X/T/

Thanks,
Liam


* Liam R. Howlett <Liam.Howlett@xxxxxxxxxx> [230714 15:56]:
> Don't set the lock to the mm lock so that the detached VMA tree does not
> complain about being unlocked when the mmap_lock is dropped prior to
> freeing the tree.
> 
> Move the destroying of the detached tree outside the mmap lock all
> together.
> 
> Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
> Signed-off-by: Liam R. Howlett <Liam.Howlett@xxxxxxxxxx>
> ---
>  mm/mmap.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/mmap.c b/mm/mmap.c
> index 7b70379a8b3e..ab6cb00d377a 100644
> --- a/mm/mmap.c
> +++ b/mm/mmap.c
> @@ -2427,7 +2427,7 @@ do_vmi_align_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
>  	unsigned long locked_vm = 0;
>  	MA_STATE(mas_detach, &mt_detach, 0, 0);
>  	mt_init_flags(&mt_detach, vmi->mas.tree->ma_flags & MT_FLAGS_LOCK_MASK);
> -	mt_set_external_lock(&mt_detach, &mm->mmap_lock);
> +	mt_detach.ma_external_lock = NULL;
>  
>  	/*
>  	 * If we need to split any vma, do it now to save pain later.
> @@ -2545,11 +2545,11 @@ do_vmi_align_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
>  	/* Statistics and freeing VMAs */
>  	mas_set(&mas_detach, start);
>  	remove_mt(mm, &mas_detach);
> -	__mt_destroy(&mt_detach);
>  	validate_mm(mm);
>  	if (unlock)
>  		mmap_read_unlock(mm);
>  
> +	__mt_destroy(&mt_detach);
>  	return 0;
>  
>  clear_tree_failed:
> -- 
> 2.39.2
>